New Dividends Appreciation data series - backtesting

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
User avatar
Simplegift
Posts: 2453
Joined: Tue Feb 08, 2011 3:45 pm
Location: Central Oregon

Re: New Dividends Appreciation data series - backtesting

Post by Simplegift » Fri Aug 11, 2017 3:00 pm

garlandwhizzer wrote:The time period we are evaluating is basically the last 30 years, a period of ever lowering interest rates/inflation and recently sluggish economic growth. Dividends as sources of income have become increasingly important ever since the bursting of the tech bubble, a time when dividends were at historical lows, I believe. The flight to dividends for yield hungry investors has been dramatic particularly over the last 8 years as interest rates have hit all-time historical lows. When you have a tail-wind your returns look nice.
Excellent point, GW. The importance of period dependency in these backtest analyses can't be stressed enough. It's one thing to evaluate dividend stocks as a source of income when 10-year Treasuries are yielding 6%, and quite another when they are yielding 2%. Context is crucial — and will continue to be so in the decades to come.
Cordially, Todd

User avatar
nedsaid
Posts: 8003
Joined: Fri Nov 23, 2012 12:33 pm

Re: New Dividends Appreciation data series - backtesting

Post by nedsaid » Fri Aug 11, 2017 3:19 pm

Okay, Vanguard High Dividend Yield Index is a measure of High Dividend and thus should relate to Large Value. How did High Dividend do as a Large Value investment? Quite well, actually.

Below is the Growth of 10,000 since 11/16/2006, date of inception for Vanguard High Dividend Yield Index. I compare it to Vanguard Value Index and the S&P 500. All funds are investor shares.

Vanguard High Dividend Yield Index $21,365
S&P 500 $21,966
Vanguard Value Index $19,251
DFA Large Cap Value I $20,526

Well, as we saw above Dividend Growth from Siamond's work and my post above seems to be a pretty good growth strategy. Problem was, Vanguard's active Dividend Growth fund did well and the Dividend Appreciation Index did relatively poorly. Dividend Growth almost matched the Growth Index whereas the Dividend Appreciation Index trailed the Growth Index badly. It was weird that active beat passive here pretty decisively. I wonder if Vanguard just picked a crummy index. There was just too much of a variance of performance. This shows that you need well constructed indexes.

My conclusion on the Value side is that High Dividend, at least during the selected time period, was a very good Value strategy. It left the Value Index in the dust and almost matched the S&P 500. Keep in mind a couple of things. First, the S&P 500 is just loaded with those successful growth stocks, and because of its market cap weighting has a bit of a Quality tilt. Second, there was a whole lot of yield chasing going on out there.

Edit: I added DFA Large Cap Value I to the analysis and you can see that it did better than Vanguard Value Index but even that still trailed the Vanguard High Dividend Yield Index. So my point about the better performance of High Dividend over plain old boring Large Cap Value still holds.

Before the financial crisis, Larry Swedroe wrote that High Dividend was a poor Value strategy. He also said that High Dividend outperformed the market over time but that was because of the underlying Value factor. Evidently, Value outperformed the market by even more. The thing was, Larry was wrong, and he was wrong because he did not account for all the yield chasing that happened after the 2008-2009 financial crisis. I believe that after the yield chasing subsides, that High Dividend will revert to being a poor Value strategy. But I don't know as evidently Larry didn't know either.

So again, in the Large Cap space, Dividend Growth almost matched the Growth Index. Both beat the S&P 500. The Vanguard Dividend Appreciation Index appears to be a bad index. Siamond's work over a longer time period showed that Dividend Growth outperformed the market handily. High Dividend also outperformed the market but by not as much as Dividend Growth. My suspicion is that a good Value index would have outperformed High Dividend over a longer time period. We saw that High Dividend outperformed in the aftermath of the 2008-2009 financial crisis, I chalk that up to yield chasing.

My conclusion, from looking at my data and then from Siamond, that dividends are not anything special on the Value side but might mean something on the Growth side. Though High Dividend outperformed Value Index since the 2008-2009 financial crisis, that was an anomaly due to yield chasing. My belief is that over long periods of time, Dividend Growth should beat the Growth index and that High Dividend will trail the Value Index. The reason that I say that Dividend Growth should beat Growth is what I see in Siamond's work and because (I think) that Dividend Growth is more highly loaded on Quality that just plain Growth. The academics say that over time Value beats the market, then surely Growth would trail the market a bit.

The slow growth and low interest rate environment we have seen since the financial crisis has favored Growth over Value. If the economy starts to rip and interest rates normalize, this would be a better environment for Value. Let's see what happens.
A fool and his money are good for business.

User avatar
siamond
Posts: 3154
Joined: Mon May 28, 2012 5:50 am

Re: New Dividends Appreciation data series - backtesting

Post by siamond » Fri Aug 11, 2017 4:04 pm

JoMoney wrote:Despite all the hype of how well a dividend based portfolio strategy would have performed, I'll point out that there are real world funds like Vanguard (and FIdelity) Equity Income Funds, and Vanguard's Dividend Growth fund, which haven't had such spectacular results and have decades of history to look at. [...] Over the past decade, we have the iShares DVY index, and the Wisdom Tree divided 'factor' funds hyped by Prof. Siegel... over the period of their existence they haven't performed anything like what the back-tested portfolios used to sell them did.
I do agree that the performance of all those dividends-oriented funds wasn't so impressive in real-life, over the past couple of decades.

Here is another telltale chart using TSM as benchmark, starting on 31-Dec-1992 (year of VDIGX's inception), and where I equated iShares DVY to TSM in the years till its inception. To focus on its real-life performance, I did NOT extend DVY with its own index (Dow Jones US Select Dividend). As a reminder, here is a brief description the various funds involved:
- VDAIX: dividend growth index fund (passive) - Nasdaq index
- VDIGX: dividend growth fund (active - focused on utilities income till 2002, then same index as VDAIX)
- VHDYX: high dividend yield index fund (passive) - FTSE index (extended backwards with MSCI)
- DVY: another index fund (passive) - DJ index follows a more complex composite strategy with elements of high-yield and elements of dividend growth

Image

Nothing really impressive in those trajectories against TSM. Note that VDIGX and DVY are burdened by an ER higher than 0.3, while the others are roughly half that. VDAIX's premium is centered on just a few years (where value surged). VDIGX and DVY Sharpe/Sortino ratios were in the same league as TSM (while VDAIX was much better).

Now if I were to represent DVY trajectory with the DJ Index filling the 1993-2003 years, you would see a much more pleasing performance, leaving all the other trajectories in the dust.

Personally, I do NOT think this is a case of theoretical backtesting of indices against the real-life of actual funds. AFAIK, those various indices seem to be defined in a pretty simple manner, and Vanguard/iShares proved without any doubt that their index funds are really good at tracking their respective indices (I did check the tracking error math on known years for those dividends funds, and it is as expected, very good).

I do think this is a case of a multi-factor strategy where its performance depends highly on the time period and where the premium is constructed over a few years of over-performance while the other years are much less exciting. To take a well-known parallel, Small/Medium Value performance over the past 10 years (compared to TSM) was a total "meh". But... Small/Medium Value truly rocked the boat in the 5 years before that. It is HIGHLY time-dependent, and the Small/Medium Value premium occurred again & again in history based on such occasional spurts of relative growth.

All this being said, I continue to be intrigued and pleasantly surprised by Dividend Growth (the strategy, not the VG active fund). The Sharpe/Sortino numbers were really good, and their behavior at times of crisis was impressive. Don't know if this will continue, but I can see why some people make that bet. Personally, I'll stick to more fundamental tilts though.

gilgamesh
Posts: 753
Joined: Sun Jan 10, 2016 9:29 am

Re: New Dividends Appreciation data series - backtesting

Post by gilgamesh » Fri Aug 11, 2017 5:29 pm

Siamond, then why didn't the Vanguard paper find the good attributes you did with the dividend-growth equites?...is it the time period tested?

User avatar
siamond
Posts: 3154
Joined: Mon May 28, 2012 5:50 am

Re: New Dividends Appreciation data series - backtesting

Post by siamond » Fri Aug 11, 2017 6:48 pm

gilgamesh wrote:Siamond, then why didn't the Vanguard paper find the good attributes you did with the dividend-growth equites?...is it the time period tested?
Well, it's a matter of perspective. A lot of material in the Vanguard paper is about debunking myths like 'dividend income matters more than total return', or 'one can safely substitute bonds by dividend stocks' (non-sense in both cases). They didn't spend much time comparing a dividend strategy to regular stocks, while it was my primary perspective.

Why are you asking? I must admit that I skimmed a bit their paper and focused mostly on understanding the strategies. Did you spot something in their findings that contradicts my findings?

gilgamesh
Posts: 753
Joined: Sun Jan 10, 2016 9:29 am

Re: New Dividends Appreciation data series - backtesting

Post by gilgamesh » Fri Aug 11, 2017 7:17 pm

siamond wrote:
gilgamesh wrote:Siamond, then why didn't the Vanguard paper find the good attributes you did with the dividend-growth equites?...is it the time period tested?
Well, it's a matter of perspective. A lot of material in the Vanguard paper is about debunking myths like 'dividend income matters more than total return', or 'one can safely substitute bonds by dividend stocks' (non-sense in both cases). They didn't spend much time comparing a dividend strategy to regular stocks, while it was my primary perspective.

Why are you asking? I must admit that I skimmed a bit their paper and focused mostly on understanding the strategies. Did you spot something in their findings that contradicts my findings?
No, I haven't read this thread or the Vanguard paper in detail, just read your conclusion and the Vanguard conclusion....sorry!

User avatar
siamond
Posts: 3154
Joined: Mon May 28, 2012 5:50 am

Re: New Dividends Appreciation data series - backtesting

Post by siamond » Fri Aug 11, 2017 10:32 pm

gilgamesh wrote:Siamond, then why didn't the Vanguard paper find the good attributes you did with the dividend-growth equites?...is it the time period tested?
Following on this comment, I looked a bit more closely to Vanguard's analysis and noticed a first mismatch. Vanguard restricted the time period to start on Jan 1st, 1997, and used the S&P 500 Dividend Aristocrats Index for the entire time period (while Vanguard's VDAIX and VDIGX both track the Nasdaq Dividend Achievers index - created mid-2006). The data series I used in previous posts spliced the S&P index and the Nasdaq index, from 1990 to 2016, which seems more realistic.

Data cover January 1, 1997, through December 31, 2016. [...] U.S. dividend growth equities are represented by the S&P 500 Dividend Aristocrats Index.

Then Vanguard had the following statement:

What is frequently overlooked, however, is that the performance of these strategies tends to vary over time and from one period to the next, and that the majority of the outperformance came from just one period: the technology stock bear market of 1999–2000, when both strategies experienced a less significant drawdown then the broad market, as shown in Figure 7. This is important because during the subsequent bear market—the 2008– 2009 global financial crisis—dividend-oriented equities did not provide the same cushion and underperformed the broader equity markets.

Well, I agree with the first point (the outperformance bit), but the point about the financial crisis is strange. First it actually seems to contradict what we can see on Figure 7 of the Vanguard paper. Next it contradicts the drawdown chart I provided in an earlier post. This being said, drawdowns are a fickle thing and using annual returns can be misleading. So I constructed monthly returns, and assembled a drawdown chart with both forms of indices (Aristocrats-only, or Composite - i.e. spliced Aristocrats/Nasdaq).

Image

What we can see is that, irrespective of the type of index being used, Dividend Growth did mitigate the damage compared to TSM during the financial crisis, although it is significantly less compelling than what the annual returns were telling us. Still, the Dividend Growth drawdown was a tad less deep and clearly narrower than TSM. So I don't think Vanguard's statement is correct, although the financial crisis was definitely frightening for every investor... As to the next crisis, time will tell if dividend growth stocks will continue to provide such mitigation cushion, hard to say.

PS. oh, and I also checked the standard deviation metrics based on monthly returns for 1990-2016, and confirmed that Dividend Growth displayed less volatility than TSM (while still acting as stocks, definitely not bonds-like).

User avatar
siamond
Posts: 3154
Joined: Mon May 28, 2012 5:50 am

Re: New Dividends Appreciation data series - backtesting

Post by siamond » Fri Aug 11, 2017 11:32 pm

While I was at it, I was curious about a monthly perspective of the corresponding growth chart. Say $1M invested in 1990. There is little question that the Dividend Growth trajectory (green line) turned out more appealing, both in terms of cumulative growth and in terms of navigating crises. Click on the image to see a larger display. Note that those are raw indices though, no expense ratios factored in.

Image

gilgamesh
Posts: 753
Joined: Sun Jan 10, 2016 9:29 am

Re: New Dividends Appreciation data series - backtesting

Post by gilgamesh » Sat Aug 12, 2017 5:38 am

Whether it be 1990 to now or 1997 to now, are we/Vanguard seriously drawing conclusions from such a narrow span? Is that reasonable?

jbolden1517
Posts: 681
Joined: Sun Jul 09, 2017 2:53 pm

Re: New Dividends Appreciation data series - backtesting

Post by jbolden1517 » Sat Aug 12, 2017 6:00 am

siamond wrote:
gilgamesh wrote:Siamond, then why didn't the Vanguard paper find the good attributes you did with the dividend-growth equites?...is it the time period tested?
Well, it's a matter of perspective. A lot of material in the Vanguard paper is about debunking myths like 'dividend income matters more than total return', or 'one can safely substitute bonds by dividend stocks' (non-sense in both cases).
Not sure I want to derail your thread here but I would argue the second isn't nonsense. Dividend stocks are kind of like a highly volatile and somewhat less income stable version of TIPS. One can easily construct a portfolio of dividend stocks with a much higher payout than the corresponding TIP payout to compensate for income volatility. At that point from an income investor's perspective it is a strictly better portfolio.

gilgamesh
Posts: 753
Joined: Sun Jan 10, 2016 9:29 am

Re: New Dividends Appreciation data series - backtesting

Post by gilgamesh » Sat Aug 12, 2017 6:02 am

siamond wrote:
What is frequently overlooked, however, is that the performance of these strategies tends to vary over time and from one period to the next, and that the majority of the outperformance came from just one period: the technology stock bear market of 1999–2000, when both strategies experienced a less significant drawdown then the broad market, as shown in Figure 7. This is important because during the subsequent bear market—the 2008– 2009 global financial crisis—dividend-oriented equities did not provide the same cushion and underperformed the broader equity markets.

Well, I agree with the first point (the outperformance bit), but the point about the financial crisis is strange. First it actually seems to contradict what we can see on Figure 7 of the Vanguard paper.
..............................
PS. oh, and I also checked the standard deviation metrics based on monthly returns for 1990-2016, and confirmed that Dividend Growth displayed less volatility than TSM (while still acting as stocks, definitely not bonds-like).
Vanguard is saying the truth in that it did not provide the same cushion in 2008-2009, but even their Figure 7 indeed show the latter half of that sentence being wrong - they did not underperform the broader equity market. Yet! Their overall statement may yet be true-

What is frequently overlooked, however, is that the performance of these strategies tends to vary over time and from one period to the next, and that the majority of the outperformance came from just one period: the technology stock bear market of 1999–2000


I am surprised with your standard deviation conclusions as well, as immediately after the above quoted paragraph, the Vanguard paper says this
The relative volatility of dividend-oriented equities has also varied considerably with time. For example, looking at rolling three-year periods, excess volatility was more than 1% higher or lower than the parent index over 58% of those periods for high-dividend-yielding equities and 56% for dividend growth equities. The relative drawdown and volatility dynamics of dividend-oriented equities are important because if the objective is to improve risk- adjusted returns, assumptions need to be made about not only future returns but also future risk
Last edited by gilgamesh on Sat Aug 12, 2017 6:23 am, edited 6 times in total.

gilgamesh
Posts: 753
Joined: Sun Jan 10, 2016 9:29 am

Re: New Dividends Appreciation data series - backtesting

Post by gilgamesh » Sat Aug 12, 2017 6:12 am

jbolden1517 wrote:
siamond wrote:
gilgamesh wrote:Siamond, then why didn't the Vanguard paper find the good attributes you did with the dividend-growth equites?...is it the time period tested?
Well, it's a matter of perspective. A lot of material in the Vanguard paper is about debunking myths like 'dividend income matters more than total return', or 'one can safely substitute bonds by dividend stocks' (non-sense in both cases).
Not sure I want to derail your thread here but I would argue the second isn't nonsense. Dividend stocks are kind of like a highly volatile and somewhat less income stable version of TIPS. One can easily construct a portfolio of dividend stocks with a much higher payout than the corresponding TIP payout to compensate for income volatility. At that point from an income investor's perspective it is a strictly better portfolio.
Relying on short span of data, or for that matter the entire past history, can indeed lead to conclusions of this sort....qualititatively, obviously non-sense

Qualitatively TIPS guarantee a certain interest rate, plus guarantees to match CPI inflation. Qualitatively, nothing else can do the same.

Now, quantitatively, you could take the only available, very recent performance of TIPS and compare its risk adjusted performance to risk adjusted return of X and show X to be better...pure non-sense...the whole point is, will it guarantee in the future, TIPS will.

Don't be blinded by numbers!

User avatar
siamond
Posts: 3154
Joined: Mon May 28, 2012 5:50 am

Re: New Dividends Appreciation data series - backtesting

Post by siamond » Sat Aug 12, 2017 7:53 am

jbolden1517 wrote:
siamond wrote:Well, it's a matter of perspective. A lot of material in the Vanguard paper is about debunking myths like 'dividend income matters more than total return', or 'one can safely substitute bonds by dividend stocks' (non-sense in both cases).
Not sure I want to derail your thread here but I would argue the second isn't nonsense. Dividend stocks are kind of like a highly volatile and somewhat less income stable version of TIPS. One can easily construct a portfolio of dividend stocks with a much higher payout than the corresponding TIP payout to compensate for income volatility. At that point from an income investor's perspective it is a strictly better portfolio.
This was just a statement for context, and I wasn't specific enough. I meant it in the sense that many people think of bonds as an income generator and a safe bucket (at least in nominal terms), and then, no form of stock can be a substitute. I do agree with you that when thinking at the portfolio level, and furthermore, when thinking of risk as something longer-term than year-to-year volatility, then the definition of a 'better' portfolio would significantly shift. Anyhoo, yes, that is a separate discussion, behavior/perception vs. cold numbers, etc.

User avatar
siamond
Posts: 3154
Joined: Mon May 28, 2012 5:50 am

Re: New Dividends Appreciation data series - backtesting

Post by siamond » Sat Aug 12, 2017 8:16 am

gilgamesh wrote:Vanguard is saying the truth in that it did not provide the same cushion in 2008-2009, but even their Figure 7 indeed show the latter half of that sentence being wrong - they did not underperform the broader equity market. Yet! Their overall statement may yet be true
I don't disagree... But as I previously explained, I would posit that in a broader historical context, factors outperformance is not a sustained thing, in the past, it was more occasional spurts of growth, which is still good to have...Combining that with the nice crisis mitigation observed in the past, this does look attractive to me. Still, there is no magic bullet (and no guarantee this will repeat itself in the future).
gilgamesh wrote:I am surprised with your standard deviation conclusions as well, as immediately after the above quoted paragraph, the Vanguard paper says this: The relative volatility of dividend-oriented equities has also varied considerably with time. [...]
Yes, I saw it, and I am not quite sure where they are coming from. The numbers are clear, standard-deviation, Sharpe & Sortino ratios. And just looking at the growth chart I provided a few posts ago should make the point pretty clear!

jbolden1517
Posts: 681
Joined: Sun Jul 09, 2017 2:53 pm

Re: New Dividends Appreciation data series - backtesting

Post by jbolden1517 » Sat Aug 12, 2017 8:19 am

gilgamesh wrote: Relying on short span of data, or for that matter the entire past history, can indeed lead to conclusions of this sort....qualititatively, obviously non-sense
Funny enough I'm going in the other direction. Consider the dividend to be a corporation's estimate of a minimum safe adjusted draw rate and assert that while there may be individual errors those can be diversified away then it makes perfect sense qualitatively to see the dividend draw as being:
a) likely to be safe
b) likely to increase with inflation
c) likely to increase at least somewhat faster than inflation as dividend companies often grow at a positive fraction of USA real growth
gilgamesh wrote: Qualitatively TIPS guarantee a certain interest rate, plus guarantees to match CPI inflation. Qualitatively, nothing else can do the same.
I would agree that nothing else is the same. But I would argue the analogy is corporate bonds :: treasuries like high dividend stocks ::TIPS . Certainly investors can use corporate bonds as a proxy for treasuries while acknowledging they are taking on credit risk. Similarly I think investors can use high dividend stocks as a proxy for TIPS while acknowledging they are taking on some economic risk.
gilgamesh wrote: Now, quantitatively, you could take the only available, very recent performance of TIPS and compare its risk adjusted performance to risk adjusted return of X and show X to be better...pure non-sense...the whole point is, will it guarantee in the future, TIPS will.
I agree TIPS are a guarantee. But ultimately how much is that guarantee worth? I think there is a spread, like in the example of corporate bonds where the guarantee is simply too expensive. It makes sense to either hold or not hold that guarantee based on the spread.

gilgamesh
Posts: 753
Joined: Sun Jan 10, 2016 9:29 am

Re: New Dividends Appreciation data series - backtesting

Post by gilgamesh » Sat Aug 12, 2017 10:57 am

siamond wrote:
Sat Aug 12, 2017 8:16 am
gilgamesh wrote:I am surprised with your standard deviation conclusions as well, as immediately after the above quoted paragraph, the Vanguard paper says this: The relative volatility of dividend-oriented equities has also varied considerably with time. [...]
Yes, I saw it, and I am not quite sure where they are coming from. The numbers are clear, standard-deviation, Sharpe & Sortino ratios. And just looking at the growth chart I provided a few posts ago should make the point pretty clear!
Sorry! I should have said, I was surprised by the Vanguard standard deviation conclusion not yours....just wanted to highlight they were different.

gilgamesh
Posts: 753
Joined: Sun Jan 10, 2016 9:29 am

Re: New Dividends Appreciation data series - backtesting

Post by gilgamesh » Sat Aug 12, 2017 11:16 am

Jbolden,

I deleted my response, as it really is taking this off topic.

User avatar
nedsaid
Posts: 8003
Joined: Fri Nov 23, 2012 12:33 pm

Re: New Dividends Appreciation data series - backtesting

Post by nedsaid » Sat Aug 12, 2017 11:34 am

siamond wrote:
Fri Aug 11, 2017 4:04 pm


All this being said, I continue to be intrigued and pleasantly surprised by Dividend Growth (the strategy, not the VG active fund). The Sharpe/Sortino numbers were really good, and their behavior at times of crisis was impressive. Don't know if this will continue, but I can see why some people make that bet. Personally, I'll stick to more fundamental tilts though.
Siamond, I do agree that Dividend Growth is a very good strategy. One reason is that growth in dividends most often relates to growth in earnings which relates to Quality. Growth in dividends also relates to consistency, which the market highly values. It seems that everyone and his brother wants consistent earnings growth and Dividend Growth seems like a good way to achieve that. Plus earnings can be manipulated to some extent, as it has been said, it is much easier to fake earnings than to fake a dividend.

What do you think about my thesis that dividends don't seem to matter on the Value side but might mean something on the Growth side? Also, does Dividend Growth beat the Growth indexes over time? The answer to the second question should answer the first.
A fool and his money are good for business.

garlandwhizzer
Posts: 1517
Joined: Fri Aug 06, 2010 3:42 pm

Re: New Dividends Appreciation data series - backtesting

Post by garlandwhizzer » Sat Aug 12, 2017 7:29 pm

Dividend growth is an attractive strategy in a number of ways. It selects solid, slowly growing companies with a long history of a growing dividend stream. That, by the way, seems to me a great definition of QUAL. DIV GR is expected to have less dramatic losses relative to both market and value in a bear market, which is what you would expect with a high quality company cranking out dividends in bad times. Like everything else, however, if this strategy becomes too popular and too much money flows that way, assets can build until they exceed available opportunities to outperform. Vanguard closed its access to Dividend Growth Fund presumably for that reason. The Vanguard Dividend Growth Fund outperformed their S&P 500 fund when it was more nimble and smaller in asset base. Ten year returns demonstrate this. However, the S&P 500 fund has outperformed the DIV GR fund over the last 1,3, and 5 years, perhaps due to two things. First too much money seeking DIV GR which seemed to provide better returns with less volatility early along. Second, as the mega cap tech darlings (FAANGs) skyrocketed, cap weighted indexes, dominated by these stocks, rode this high-growth-mega-cap wave upward, while DIV GR, whose holdings are more old economy, reliable, slow growing, and less volatile did not. The point is that results of this DIV GR approach are period dependent and also subject to capacity restraints if too much money flows that way, which it tends to happen to all successful funds and approaches, particularly those that offer the promise of higher returns with lower risk. When and if this high growth-mega-cap tech boom plays out and/or reverses, DIV GR might once again be expected to outperform, especially if it has lost some of its popularity by then.

Garland Whizzer

Nate79
Posts: 977
Joined: Thu Aug 11, 2016 6:24 pm
Location: Portland, OR

Re: New Dividends Appreciation data series - backtesting

Post by Nate79 » Sat Aug 12, 2017 7:56 pm

It's surprising to me that DIV GR yield is very close to TSM. I would have guessed it would be higher considering it is a dividend strategy.

User avatar
siamond
Posts: 3154
Joined: Mon May 28, 2012 5:50 am

Re: New Dividends Appreciation data series - backtesting

Post by siamond » Sun Aug 13, 2017 12:38 pm

Nate79 wrote:
Sat Aug 12, 2017 7:56 pm
It's surprising to me that DIV GR yield is very close to TSM. I would have guessed it would be higher considering it is a dividend strategy.
I suspect this has to do with the fact that Dividend Growth (e.g. VDAIX) is about growing dividends, and not about High Dividend Yield (e.g. VHDYX), which does display a SEC Yield one full point over TSM and VDAIX - at the time of writing.

MIpreRetirey
Posts: 616
Joined: Fri Sep 06, 2013 12:35 pm

Re: New Dividends Appreciation data series - backtesting

Post by MIpreRetirey » Sun Aug 13, 2017 2:28 pm

siamond wrote:
Sun Aug 13, 2017 12:38 pm
Nate79 wrote:
Sat Aug 12, 2017 7:56 pm
It's surprising to me that DIV GR yield is very close to TSM. I would have guessed it would be higher considering it is a dividend strategy.
I suspect this has to do with the fact that Dividend Growth (e.g. VDAIX) is about growing dividends, and not about High Dividend Yield (e.g. VHDYX), which does display a SEC Yield one full point over TSM and VDAIX - at the time of writing.
Don't consider this authorative, but I checked just 2 data points for S&P500 and Div.Apprec.Idx, each, for div income.

1st, the yield that you mention is earnings yield?

2nd, I compared both price expansion and Tot.Ret. expansion since 03/2009 for VFINX and VDAIX, each.
Both price and TR have VFINX ahead by about 25% over VDAIX since 03/13/2009. Which says they both had the same P/E expansion.

Next, 2 point dividends paid points after 5 years, for each fund: $.61 grew to $.96 for VFINX.
$.12 grew to $.20 for VDAIX. The 2 data points were 06/2012 and 06/2017, for both funds.

So, although the earnings didn't increase any faster( was the same) for VDAIX, the Div. payout grew more.


Edit: Apologies. TTM and Sec-30 day Yields are for income/dividends, not earnings. Give me time and I'll post something else dumb. Details, details. :oops:

So, redoing this math using matching dates for div and TR:
05/2012 to 06/2017:
VFINX/VDAIX TR growth was 19% more for VFINX.
VFINX/VDAIX price growth was 23% more for VFINX.
Which says that the P/E P/D expansion of VFINX was higher than for VDAIX. Or VFINX's yield has come down relative to VDAIX.
Tried to fix this!

Next, 2 point dividends paid points after 5 years, for each fund: $.61 grew to $.96 for VFINX.
$.12 grew to $.20 for VDAIX. The 2 data points were 06/2012 and 06/2017, for both funds.
Which was a 67% increase for VDAIX and a 57% increase for VFINX.


(BTW, siamond. Looks like I can export images to desktop from my Simba sheets, and post, labels included, without the usual screenshot/crop in MSpaint 1st. Nice!)( Great thread. Considering adding some of my own data series to use.)

User avatar
siamond
Posts: 3154
Joined: Mon May 28, 2012 5:50 am

Re: New Dividends Appreciation data series - backtesting

Post by siamond » Sun Aug 13, 2017 4:11 pm

nedsaid wrote:
Sat Aug 12, 2017 11:34 am
What do you think about my thesis that dividends don't seem to matter on the Value side but might mean something on the Growth side? Also, does Dividend Growth beat the Growth indexes over time? The answer to the second question should answer the first.
Well... To be honest, I am not terribly convinced! :wink:

First, I don't quite see the relation between Dividend Growth (which is about growing dividend yields) and Growth (as in growth vs. value; i.e. a classification based on measures of the company's valuation via price/book and other criteria). It's just a case of an unfortunate overload of the word 'growth'. Next here is a telltale chart (to your request) introducing Large-Cap-Value (LCV) and Large-Cap-Growth (LGC). Click on the pic for a larger display.

Image

As you can, LCV and LCG switched places and ended up remarkably on par with TSM. There was just no large-value effect in the past 30 years or so (only small/mid value made a difference). Now, as to patterns, I really don't see much linking LCV or LCG to the dividend funds trajectories. I took a quick look at Mid-Cap Growth (MCG) as well, and same outcome.

Finally, if you check the correlation table from this post, you'll see that both dividend strategies displayed more correlation with LCV than LCG. So... no, I don't think Dividend Growth had much to do with the 'Growth side'.

gilgamesh
Posts: 753
Joined: Sun Jan 10, 2016 9:29 am

Re: New Dividends Appreciation data series - backtesting

Post by gilgamesh » Sun Aug 13, 2017 4:16 pm

Siamond

So, the good attributes of dividend growth seen in the span tested cannot be explained away by any factor influence? It's unique to the dividend growth group? Or is this not correct?

This is a silly question, any parameters that could predict the period where dividend-growth will outperform TSM? I understand, there's no need to do this, as if the span tested will continue to repeat then just holding dividend growth over TSM in perpetuity will deliver better returns and lower volatility especially nice damping during market crashes, right??

User avatar
nedsaid
Posts: 8003
Joined: Fri Nov 23, 2012 12:33 pm

Re: New Dividends Appreciation data series - backtesting

Post by nedsaid » Sun Aug 13, 2017 4:34 pm

Siamond, thank you for your work. At least in the Large Cap space, it appears there is no correlation between High Dividend and Value, or between Dividend Growth and Growth. Also, Value and Growth over 30 years tied the Total Stock Market Index. Keep in mind, I am not aware of a "Quality Fund" run by Vanguard or DFA that we can use as a benchmark against Dividend Growth. So pretty much, I used the Growth Index as a proxy for a "Quality Index." So I don't know, at least from your work, it appears that I was wrong about almost everything I said. Oh well, nothing new. I am right about anything around here only about every six months or so! :wink:

Vanguard splits the market in two and calls 1/2 Value and the other 1/2 Growth. The academics define Value as 30% of the market and not 50%. My guess is that Larry Swedroe would say that the Vanguard Value Index does not load very well on Value. As they say, when at first you try and don't succeed, blame the benchmarks. My guess is that Larry would say that the benchmarks we are using are flawed, the "Bad Index" argument.

So pretty much what I get out of Siamond's excellent work is that Small/Mid-Cap Value and Dividend Growth are mighty fine strategies and that High Dividend Yield beats the market by a little. My guess is that the Value and Growth Indexes used here are pretty useless for factor analysis. Otherwise, Larry would be just plain wrong and that dividends in themselves would seem to be a factor. I kind of like the "blame the indexes" argument.
A fool and his money are good for business.

User avatar
siamond
Posts: 3154
Joined: Mon May 28, 2012 5:50 am

Re: New Dividends Appreciation data series - backtesting

Post by siamond » Sun Aug 13, 2017 5:12 pm

nedsaid wrote:
Sun Aug 13, 2017 4:34 pm
At least in the Large Cap space, it appears there is no correlation between High Dividend and Value, or between Dividend Growth and Growth.
gilgamesh wrote:
Sun Aug 13, 2017 4:16 pm
So, the good attributes of dividend growth seen in the span tested cannot be explained away by any factor influence? It's unique to the dividend growth group? Or is this not correct?
Hm, I didn't quite say that, did I? I did point out that Value and Min. Volatility were fairly strongly correlated with both dividend strategies (cf. the -annual- correlation matrix in this post). Quality is much less clear though. I will re-run correlation numbers with factor series on a monthly basis though, to get a better read on this. And I will also try to better understand how Vanguard did its 'factor attribution' decomposition, I suspect my logic is too simplistic. You guys got me curious... :wink:

gilgamesh wrote:
Sun Aug 13, 2017 4:16 pm
This is a silly question, any parameters that could predict the period where dividend-growth will outperform TSM? I understand, there's no need to do this, as if the span tested will continue to repeat then just holding dividend growth over TSM in perpetuity will deliver better returns and lower volatility especially nice damping during market crashes, right??
I am only sharing my observations about the past, I will let you speculate about the future. Please consider that 32 years (with two major crises) isn't that long of a time span to reach conclusions, and that the human brain tends to see patterns a little too fast for its own sake! Notably when influenced by pre-conceived ideas (aka confirmation bias). Still, dividend growth did display a nice track record. I'd be quite curious to see what will happen during the next crisis, actually. And in a period of time where interest rates will probably be on the rise.
nedsaid wrote:
Sun Aug 13, 2017 4:34 pm
So pretty much what I get out of Siamond's excellent work is that Small/Mid-Cap Value and Dividend Growth are mighty fine strategies and that High Dividend Yield beats the market by a little.
If you were to reword this sentence using past tenses, then yes, this seems like a good bottomline.

User avatar
nedsaid
Posts: 8003
Joined: Fri Nov 23, 2012 12:33 pm

Re: New Dividends Appreciation data series - backtesting

Post by nedsaid » Sun Aug 13, 2017 9:49 pm

siamond wrote:
Sun Aug 13, 2017 5:12 pm
nedsaid wrote:
Sun Aug 13, 2017 4:34 pm
So pretty much what I get out of Siamond's excellent work is that Small/Mid-Cap Value and Dividend Growth are mighty fine strategies and that High Dividend Yield beats the market by a little.
If you were to reword this sentence using past tenses, then yes, this seems like a good bottomline.
Siamond, I use present tense. If the success of these strategies are indeed factor based, I would expect that success to continue. I say that because of my belief that factors are behaviorally based and also upon my belief that human nature and behavior doesn't change over time. You are right, no one knows for sure what will happen in the future. As they say, go big or go home. So if I am wrong, I may as well be very wrong. :wink:
A fool and his money are good for business.

User avatar
siamond
Posts: 3154
Joined: Mon May 28, 2012 5:50 am

Re: New Dividends Appreciation data series - backtesting

Post by siamond » Sun Aug 13, 2017 11:29 pm

nedsaid wrote:
Sun Aug 13, 2017 9:49 pm
Siamond, I use present tense. If the success of these strategies are indeed factor based, I would expect that success to continue. I say that because of my belief that factors are behaviorally based and also upon my belief that human nature and behavior doesn't change over time. You are right, no one knows for sure what will happen in the future. [...]
I use various tilts myself, so you can guess that, as a personal opinion, I would tend to agree, unless proven otherwise (LCV being a good example!). But I try my best to stay factual in this kind of analysis, hence my use of the past tense (a small Nisiprius is sitting on my shoulder, mercilessly chastising me otherwise! :wink:).

lazyday
Posts: 2876
Joined: Wed Mar 14, 2007 10:27 pm

Re: New Dividends Appreciation data series - backtesting

Post by lazyday » Mon Aug 14, 2017 1:21 am

nedsaid wrote:
Sun Aug 13, 2017 9:49 pm
If the success of these strategies are indeed factor based, I would expect that success to continue.
Some value indexes and funds choose a single measure of value, such as price to book or dividend yield. Others use several measures.

IMO it's possible for value in general to become so popular that it's overpriced, and nearly as expensive as the broad market. In that case, the value premium could become negative.

I think there is a greater risk of something like this if you choose a single measure of value. Perhaps high yield could become so popular that the expected premium is negative, while value in general still has a tiny (positive) premium.

Also, the behavioral explanations for some factors are more believable than others. I'm not convinced that quality or low volatility will persist. Though they might still be worthwhile, as long as the negative premia aren't too punishing.

anil686
Posts: 628
Joined: Thu May 08, 2014 12:33 pm

Re: New Dividends Appreciation data series - backtesting

Post by anil686 » Mon Aug 14, 2017 6:17 am

siamond wrote:
Fri Aug 11, 2017 11:32 pm
While I was at it, I was curious about a monthly perspective of the corresponding growth chart. Say $1M invested in 1990. There is little question that the Dividend Growth trajectory (green line) turned out more appealing, both in terms of cumulative growth and in terms of navigating crises. Click on the image to see a larger display. Note that those are raw indices though, no expense ratios factored in.

Image
Thanks for the great info in this thread and really enjoyed it. When you used Dividend growth above - which index or rule did you use? was it Aristocrats with the 25 years of increasing dividends, the Nasdaq Achievers index (VG with 10 years) or Morningstar's Index - the five years of growing dividends?

Thanks in advance!

User avatar
siamond
Posts: 3154
Joined: Mon May 28, 2012 5:50 am

Re: New Dividends Appreciation data series - backtesting

Post by siamond » Mon Aug 14, 2017 9:30 am

anil686 wrote:
Mon Aug 14, 2017 6:17 am
When you used Dividend growth above - which index or rule did you use? was it Aristocrats with the 25 years of increasing dividends, the Nasdaq Achievers index (VG with 10 years) or Morningstar's Index - the five years of growing dividends?
On this chart, I was using a composite approach, the S&P 500 Aristocrats spliced with the Nasdaq Achievers (starting from 2007).

anil686
Posts: 628
Joined: Thu May 08, 2014 12:33 pm

Re: New Dividends Appreciation data series - backtesting

Post by anil686 » Mon Aug 14, 2017 10:54 am

siamond wrote:
Mon Aug 14, 2017 9:30 am
anil686 wrote:
Mon Aug 14, 2017 6:17 am
When you used Dividend growth above - which index or rule did you use? was it Aristocrats with the 25 years of increasing dividends, the Nasdaq Achievers index (VG with 10 years) or Morningstar's Index - the five years of growing dividends?
On this chart, I was using a composite approach, the S&P 500 Aristocrats spliced with the Nasdaq Achievers (starting from 2007).
Thanks!

Also, when you plotted the different dividend strategies in a tell tale with TSM back to 1992 - how did you represent both the Aristocrats and the Achievers? Were there indexes (without funds) available for these or was this an estimate based on modeling? I found this fascinating - especially since the actively managed dividend growth fund by VG has done well recently but on the chart from 1992, not so much while the dividend achievers index has not done as well recently, but as you pointed out - achieved a tremendous amount of it's outperformance in the 2000-2002 period.

Thanks again!!!

User avatar
siamond
Posts: 3154
Joined: Mon May 28, 2012 5:50 am

Re: New Dividends Appreciation data series - backtesting

Post by siamond » Mon Aug 14, 2017 11:42 am

anil686 wrote:
Mon Aug 14, 2017 10:54 am
Also, when you plotted the different dividend strategies in a tell tale with TSM back to 1992 - how did you represent both the Aristocrats and the Achievers? Were there indexes (without funds) available for these or was this an estimate based on modeling?
As I said, I spliced the two indices. So I used the total returns from Aristocrats from 1990 to 2006, and Nasdaq Achievers from 1997 to 2016. The latter is a more modern index, cap-weighted, and the one the corresponding Vanguard funds track, so I used as much history I could from it. But then to go back further in history, we have to rely on the Aristocrats (which is equal-weighted, so it's a bit different, although the general strategy is broadly similar), and this index has a much longer history, back to 1990. This is the best we can do, historical records are what they are... No modeling involved here, all numbers are available on Morningstar.

gilgamesh
Posts: 753
Joined: Sun Jan 10, 2016 9:29 am

Re: New Dividends Appreciation data series - backtesting

Post by gilgamesh » Mon Aug 14, 2017 11:47 am

nedsaid wrote:
Sun Aug 13, 2017 9:49 pm
siamond wrote:
Sun Aug 13, 2017 5:12 pm
nedsaid wrote:
Sun Aug 13, 2017 4:34 pm
So pretty much what I get out of Siamond's excellent work is that Small/Mid-Cap Value and Dividend Growth are mighty fine strategies and that High Dividend Yield beats the market by a little.
If you were to reword this sentence using past tenses, then yes, this seems like a good bottomline.
Siamond, I use present tense. If the success of these strategies are indeed factor based, I would expect that success to continue. I say that because of my belief that factors are behaviorally based and also upon my belief that human nature and behavior doesn't change over time. You are right, no one knows for sure what will happen in the future. As they say, go big or go home. So if I am wrong, I may as well be very wrong. :wink:
Can you link some threads or articles or suggest books that could elaborate on this...if this is true, I want to study that more...thanks!..sry! for the slight hijack.

User avatar
nedsaid
Posts: 8003
Joined: Fri Nov 23, 2012 12:33 pm

Re: New Dividends Appreciation data series - backtesting

Post by nedsaid » Mon Aug 14, 2017 1:07 pm

Gilgamesh, there have been many threads on factor investing. When Larry Swedroe used to hold court over the factor threads, he felt that factors, particularly Value, were a combination of a behavioral story and a risk story. He often said that while Value wasn't a free lunch, it was a free stop at the dessert tray. One reason I believe the behavioral explanation is that factors are persistent and pervasive. Read through those threads, and you will see the usual suspects there. This has been hashed over many times.
A fool and his money are good for business.

gilgamesh
Posts: 753
Joined: Sun Jan 10, 2016 9:29 am

Re: New Dividends Appreciation data series - backtesting

Post by gilgamesh » Mon Aug 14, 2017 1:16 pm

nedsaid wrote:
Mon Aug 14, 2017 1:07 pm
Gilgamesh, there have been many threads on factor investing. When Larry Swedroe used to hold court over the factor threads, he felt that factors, particularly Value, were a combination of a behavioral story and a risk story. He often said that while Value wasn't a free lunch, it was a free stop at the dessert tray. One reason I believe the behavioral explanation is that factors are persistent and pervasive. Read through those threads, and you will see the usual suspects there. This has been hashed over many times.
Thanks!...will do.

User avatar
siamond
Posts: 3154
Joined: Mon May 28, 2012 5:50 am

Re: New Dividends Appreciation data series - backtesting

Post by siamond » Mon Aug 14, 2017 7:16 pm

One last thing... I assembled more extensive monthly data series, notably data series capturing factors (momentum, quality, min volatility, value) thanks to corresponding MSCI indices. Now when comparing correlations (e.g. dividend growth against value), the outcome is quite blurred by the fact that both data series are themselves strongly correlated with the total US market (TSM).

It then struck me that it could be interesting to compute premium data series (e.g. dividend growth minus TSM; quality minus TSM; etc; and also Small-Caps minus TSM to capture the size premium; I also added an MCV and an SCV premium while I was at it). By 'minus', I mean a geometric difference (i.e. (1+X)/(1+Y)-1).

Once we have those 'premium' data series, we can check how the premiums of real-life asset classes (dividend growth, dividend high yield, or others like REITs) correlate to the premiums displayed by the 'factors', and then we should have a clearer view of which factors influence the most the trajectory of a given asset class. See below. Yes, I know, a lot of small numbers, click on the image to see a bigger display.

Image

And... this premium perspective didn't tell me much more than what was observed with simpler correlations:
- both dividend growth and dividend high yield have a significant correlation to min. volatility and value
- I continue to not see any significant correlation with quality (contrary to the Vanguard study)
- there is a negative correlation with size, which isn't surprising, dividend strategies tend to hone on large-caps

Ok, I am not quite sure how meaningful (or actionable) this is all is. Personally, I would never choose an asset class based on indirect metrics like factor exposure and so on. Still, I tried to read more about the way the pros compute factor exposures (e.g. this link), but this quickly gave me a headache, none of it being intuitive - contrary to a premium analysis). So I think I'm going to stop the investigation at this point... And yes, the various data series we discussed will be available in the next update of the Simba spreadsheet.

User avatar
nedsaid
Posts: 8003
Joined: Fri Nov 23, 2012 12:33 pm

Re: New Dividends Appreciation data series - backtesting

Post by nedsaid » Mon Aug 14, 2017 7:30 pm

gilgamesh wrote:
Mon Aug 14, 2017 1:16 pm
nedsaid wrote:
Mon Aug 14, 2017 1:07 pm
Gilgamesh, there have been many threads on factor investing. When Larry Swedroe used to hold court over the factor threads, he felt that factors, particularly Value, were a combination of a behavioral story and a risk story. He often said that while Value wasn't a free lunch, it was a free stop at the dessert tray. One reason I believe the behavioral explanation is that factors are persistent and pervasive. Read through those threads, and you will see the usual suspects there. This has been hashed over many times.
Thanks!...will do.
Go through Larry Swedroe's posts, that will help. There were factor threads, dividend threads, value threads and the usual suspects battled each other. Larry told us we were all wet. :wink: I still have to get Larry's recent book on factor investing.
A fool and his money are good for business.

Longtermgrowth
Posts: 314
Joined: Thu Nov 26, 2015 1:59 pm

Re: New Dividends Appreciation data series - backtesting

Post by Longtermgrowth » Tue Aug 15, 2017 3:36 am

Very nice thread. Appreciate the work, siamond. I'm curious where everyone would put the Dow Jones U.S. Dividend 100 Index (SCHD, Schwab U.S. Dividend Equity ETF tracks this) compared to Vanguard's High Dividend and Dividend Appreciation. I've seen it mentioned a number of times that SCHD could be viewed as a hybrid of the two Vanguard funds.

lazyday
Posts: 2876
Joined: Wed Mar 14, 2007 10:27 pm

Re: New Dividends Appreciation data series - backtesting

Post by lazyday » Tue Aug 15, 2017 5:32 am

Longtermgrowth wrote:
Tue Aug 15, 2017 3:36 am
I'm curious where everyone would put the Dow Jones U.S. Dividend 100 Index (SCHD
If there's enough data, you might be able to use regression to find the factor exposure. Just from a look at the methodology, it seems like a multifactor fund with exposure to quality, value, and low vol.

Here's some of the methodology from page 24, leaving out many details:
http://us.spindices.com/documents/metho ... nload=true
2,500 largest U.S. stocks .... screens ... Minimum 10 consecutive years of dividend payments

.... ranked in descending order by indicated annual dividend yield ... The top half of securities based on this ranking are eligible ....

1. The eligible securities are ranked by each of four fundamentals-based characteristics:
• Cash flow to total debt
• Return on equity
• Indicated dividend yield
• Five-year dividend growth rate
.... The 100 top-ranked stocks by the composite score are selected to the index ... buffer rules ....

Weightings ... modified market capitalization ... No single stock can represent more than 4.5% of the index and no single industry ... more than 25%
So we have some value, by selecting top half of yield, and again ranking those partly based on yield.
We have some quality, by ranking partly on cash flow/debt and ROE. Div growth may also be related to quality.
The quality rankings probably give exposure to low vol. For example, ability to service debt makes companies less risky.

Unlike the other dividend "growth" indexes in this thread, this index rewards high growth, not just any nominal growth above 0, even if growth after inflation is negative. But it does allow dividend cuts (and zero growth) unlike the other dividend growth indexes.

It is high yield, but probably not as high as most funds that only target yield.

EDIT: added info on weighting to explain "modified". Note that the 4.5% cap makes it a bit like an equal weighted fund such as BRLIX.
Last edited by lazyday on Tue Aug 15, 2017 10:11 am, edited 1 time in total.

lazyday
Posts: 2876
Joined: Wed Mar 14, 2007 10:27 pm

Re: New Dividends Appreciation data series - backtesting

Post by lazyday » Tue Aug 15, 2017 8:52 am

I don't know if there's enough data for these results to be useful, but a regression analysis of SCHD shows a negative value load, moderate profitability, and high conservative-investment. R-sq is only 91.5% even with 5 factors.

This seems to be the P/B (or high B/P minus low) version of value, which might help explain the poor value loading. I believe that quality companies often have high P/B, even if other price multiples are reasonable. Quality has high ROI, ROE, ROA, so we expect high earnings to book, and high price to book. A broader measure of value might produce different results. Though I don't really know anything about factor analysis.

https://www.portfoliovisualizer.com/fac ... ssion=true

User avatar
siamond
Posts: 3154
Joined: Mon May 28, 2012 5:50 am

Re: New Dividends Appreciation data series - backtesting

Post by siamond » Tue Aug 15, 2017 10:08 am

lazyday wrote:
Tue Aug 15, 2017 8:52 am
I don't know if there's enough data for these results to be useful, but a regression analysis of SCHD shows a negative value load, moderate profitability, and high conservative-investment. R-sq is only 91.5% even with 5 factors. [...]
https://www.portfoliovisualizer.com/fac ... ssion=true
Oh, that is really cool... PortfolioVisualizer did all the complicated 'factor exposure' math and created a nice user interface around it. I had no idea. Thank you for sharing. Unfortunately, for most of the funds that we're discussing, they didn't exist for very long, so this isn't terribly significant, but the tool itself is really cool. I'll ask the PV author if there is a way to extend his tool to index data series as this would be the real key here.

lazyday
Posts: 2876
Joined: Wed Mar 14, 2007 10:27 pm

Re: New Dividends Appreciation data series - backtesting

Post by lazyday » Tue Aug 15, 2017 10:14 am

Surprised you haven't seen it, but glad to help!

User avatar
siamond
Posts: 3154
Joined: Mon May 28, 2012 5:50 am

Re: New Dividends Appreciation data series - backtesting

Post by siamond » Tue Aug 15, 2017 10:45 am

Longtermgrowth wrote:
Tue Aug 15, 2017 3:36 am
Very nice thread. Appreciate the work, siamond. I'm curious where everyone would put the Dow Jones U.S. Dividend 100 Index (SCHD, Schwab U.S. Dividend Equity ETF tracks this) compared to Vanguard's High Dividend and Dividend Appreciation. I've seen it mentioned a number of times that SCHD could be viewed as a hybrid of the two Vanguard funds.
Welcome. Yes, the Dow Jones dividends-oriented indices have more complex filters and should be considered as an hybrid strategy. I didn't look much at SCHD and its index (besides the information lazyday already provided).

I did look at iShares Dow Jones Select Dividend Index (DVY), which tracks the related index (duh!), and this is definitely an hybrid (methodology document can be found here). Its index actually displayed quite a remarkable performance since 1992 (better than pure dividend growth funds; and trajectory quite similar to MCV, interestingly enough). Choices, choices...

User avatar
siamond
Posts: 3154
Joined: Mon May 28, 2012 5:50 am

Re: New Dividends Appreciation data series - backtesting

Post by siamond » Tue Aug 15, 2017 4:05 pm

lazyday wrote:
Tue Aug 15, 2017 10:14 am
Surprised you haven't seen it, but glad to help!
I never really looked at this factor exposure mathematical stuff until this thread... Always a learning process... :P

Ok, the PV author explained how to import monthly data series with my login, which allowed me to run the factor regression analysis on index data series. So I imported the six indices we've been discussing so far, and here is the outcome (Rm-Rf is market; SMB is size; HML is value; MOM is momentum; QMJ is Quality; low volatility doesn't seem to be available on PV):

Image

Then we're back to the Vanguard finding that the Dividend Growth strategies (Aristocrats & Achievers) have more quality exposure than the Dividend High Yield approach. And the 'hybrid' strategies from Dow Jones are more value-y. I am a tad baffled that my naive way of proceeding (correlation between premiums) doesn't find something similar, but I'll readily admit that I don't master this stuff, and I have no doubt that the PV author (who is a really smart guy, by the way) does! I am still not sure how this is actionable, but well, we have the proper data available for those of you who are interested!

lazyday
Posts: 2876
Joined: Wed Mar 14, 2007 10:27 pm

Re: New Dividends Appreciation data series - backtesting

Post by lazyday » Tue Aug 15, 2017 4:55 pm

siamond wrote:
Tue Aug 15, 2017 4:05 pm
Ok, the PV author explained how to import monthly data series with my login
Wow, that's great!
Then we're back to the Vanguard finding that the Dividend Growth strategies (Aristocrats & Achievers) have more quality exposure than the Dividend High Yield approach.
I would really expect that. Since the Aristocrats and Achievers avoided dividend cuts for 25 and 10 years, they probably have some earnings stability and/or low debt and/or high ROx. For example with low ROA, you might get in trouble in bad years with a sales drop if you can't cover fixed costs. I don't know the accounting here, so if I'm off someone please correct this.

Really I'm surprised the two "High Dividend Yield" indexes have the QMJ ratings that they do, and wonder where it comes from. From SMB we see they're largecap or megacap funds, and large is associated with quality, but I don't know how strongly.
I am still not sure how this is actionable, but well, we have the proper data available for those of you who are interested!
I'm looking at the bottom one, DJ 100. Is the 84% Rsq high enough to trust the factor loadings?

By the way, it might be interesting to see QUAL's dataset in there too, if it's easy to do. To compare a straight quality index.

User avatar
siamond
Posts: 3154
Joined: Mon May 28, 2012 5:50 am

Re: New Dividends Appreciation data series - backtesting

Post by siamond » Tue Aug 15, 2017 6:37 pm

lazyday wrote:
Tue Aug 15, 2017 4:55 pm
By the way, it might be interesting to see QUAL's dataset in there too, if it's easy to do. To compare a straight quality index.
Yes, you read my mind, this was indeed my next step... I uploaded the data series for MSCI Momentum, Quality and Value indices. And I added a few Vanguard funds, just for fun. See below. Personally, this leaves me shaking my head a little bit. My guess is that there are way too many definitions of value, quality, momentum and so on for such exposure math to be truly meaningful... I used the AQR 4-factors model, by the way (as this is the only one with a 'quality' metric).

Image

lazyday
Posts: 2876
Joined: Wed Mar 14, 2007 10:27 pm

Re: New Dividends Appreciation data series - backtesting

Post by lazyday » Wed Aug 16, 2017 5:02 am

siamond wrote:
Tue Aug 15, 2017 6:37 pm
I uploaded the data series for MSCI Momentum, Quality and Value indices.
Thanks, nice to see these too.
Personally, this leaves me shaking my head a little bit. My guess is that there are way too many definitions of value, quality, momentum and so on for such exposure math to be truly meaningful
Just from my limited understandings:

As a sector neutral fund, QUAL won't have very high loading. It has market weighting of low quality sectors such as financials and consumer discretionary. But I'm not aware of any other single factor index fund of reasonable ER that directly targets quality. SPHQ is .29% ER vs QUAL .15 (and SCHD .07, thanks Longtermgrowth).

The Vanguard SV fund used an S&P index in the first years, and the S&P profit screens might have added a bit of quality loading.
Some have posted that value is strongest in SV. I don't know if small quality companies are higher quality than large quality companies, but large companies generally tend to be higher quality than small.
It seems crazy to me that SV has higher quality loading than largecap QUAL. It's also higher than the largecap Nasdaq Achievers, though that doesn't directly target quality.
I used the AQR 4-factors model, by the way (as this is the only one with a 'quality' metric)
5 factor, right? I wan't able to select AQR 5 factor when I tried SCHD.

The FF factors include Profitability which is a type of quality, though not as broad as AQR's QMJ which uses "various measures of profitability, growth, safety and payout".

FF also includes (low) Investment, which I believe is related to quality. It's a bit confusing to me, as I'd think you would have some companies that are able to produce high earnings with low investment (Buffett quality before he bought a railroad) but also some unprofitable struggling companies that can't afford investment? I haven't read a paper about CMA yet.

By the way, SMLF is a smallcap fund using an MSCI index to target value, quality, momentum, and low size. But that's not a request, just if you're interested in comparing it to the SV fund's quality exposure. I believe Larry was happy with this family of multifactor funds.

User avatar
siamond
Posts: 3154
Joined: Mon May 28, 2012 5:50 am

Re: New Dividends Appreciation data series - backtesting

Post by siamond » Wed Aug 16, 2017 8:19 am

lazyday wrote:
Wed Aug 16, 2017 5:02 am
I used the AQR 4-factors model, by the way (as this is the only one with a 'quality' metric)
5 factor, right? I wan't able to select AQR 5 factor when I tried SCHD.
Yes, same thing here, so I settled for AQR 4 factors. Apparently a limitation of PV's factor analysis.

User avatar
nedsaid
Posts: 8003
Joined: Fri Nov 23, 2012 12:33 pm

Re: New Dividends Appreciation data series - backtesting

Post by nedsaid » Wed Aug 16, 2017 10:41 am

siamond wrote:
Tue Aug 15, 2017 6:37 pm
My guess is that there are way too many definitions of value, quality, momentum and so on for such exposure math to be truly meaningful... I used the AQR 4-factors model, by the way (as this is the only one with a 'quality' metric).
Yes, all of this seems to be in the eye of the beholder to some degree. I remember Larry Swedroe telling the Benjamin Graham practitioners that they were all wet about their definition of Value. There are definitely differences of opinion.

By the way, I read the Vanguard white paper on factor investing, it was quite interesting. Below is my main takeaway from the paper:
The performance of dividend-oriented strategies can be largely explained by their exposure to a small number of equity factors: value and lower volatility for high-dividend-yielding equities, and lower volatility and quality for dividend growth equities.
This reinforces the common belief that dividends are a cushion in bear markets.
A fool and his money are good for business.

Post Reply